This is the mail archive of the libstdc++@gcc.gnu.org mailing list for the libstdc++ project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: MIPS libstdc++-v3 x86-x-mips-elf doesn't build



> I'm trying to build a x86-x-mips-elf toolchain for testing MIPS patches
> using gcc + binutils + newlib + gdb. 

OK

> Since this patch by HJL I have been unable to build libsdtc++ for a
> x86-x-mips-elf toolchain.
> 
> 2001-07-25  H.J. Lu <hjl@gnu.org>
> 
>         * configure.target (cpu_include_dir): Set to `config/cpu/mips'
>         for mips.

This part is fine.

>         * config/cpu/mips/bits/atomicity.h: New.

This part is not.

#include <sgidefs.h>

typedef int _Atomic_word;

#if (_MIPS_ISA >= _MIPS_ISA_MIPS2)

First of all, sgidefs should not be included, since this is a target for 
explictly non-IRIX systems (irix systems use os/irix atomicity.h 
headers). So, this part needs to go. Also, the _MIPS_ISA bits probably 
need to be replaced with something target-specific, or sizeof(int) >= 4 or 
something. 

H.J., what is the difference between the ISA bits in this file, as 
checked in? What's your intention here?

Sorry, I didn't notice this first time around.

> What gives? I'm not targeting a SGI target so I don't expect to need
> any SGI header files.

Right.

> The other thing is I'm not targeting IRIS5, IRIS6 or Linux so both
> _MIPS_ISA and _MIPS_ISA_MIPS1 are undefined. This leads to a problem
> because then the preprocessor test (_MIPS_ISA >= _MIPS_ISA_MIPS1) then
> evaluates to true and we think we are targeting a MIPS2 or greater ISA.

Ok. Is there a way to do this with sizeof or something?

> For the moment I've hacked up my atomicity.h so that I can build libstdc++v3

What did you do? You can also use the generic atomicity.h header for the 
time being.

-benjamin


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]